Predictive model for prolonged length of hospital stay in patients with osteoporotic femoral neck fracture: A 5-year retrospective study.

Prolonged length of stay (LOS) in osteoporotic femoral neck fracture patients increased the hospital care cost and demonstrated in-hospital complications. This study aimed to develop an ease-of use predictive model of prolonged LOS in osteoporotic femoral neck fracture patients. In this 5-year retrospective study, the medical charts of 255 patients admitted to hospital with an osteoporotic femoral neck fracture resulting from a simple fall from January 2014 to December 2018 were reviewed. Multivariable fractional polynomials (MFP) algorithms was applied to develop the predictive model from candidate predictors of prolonged LOS. The discrimination performance of predictive model was evaluated using the receiver operating characteristic curve (ROC). Internal validity was assessed using bootstrapping. From 289 patients who were hospitalized with an osteoporotic fracture of femoral neck throughout this study, 255 (88%) fulfilled the inclusion criteria. There was 54.90% (140 of 255 patients) of patients who had prolonged LOS. The predictors of the predictive model were age, BMI, ASA score class 3 or 4, arthroplasty and time from injury to surgery. The area under ROC curve of the model was 0.83 (95% confidence interval 0.77-0.88). Internal validation with bootstrap re-sampling revealed an optimism of -0.002 (range -0.300-0.296) with an estimated shrinkage factor of 0.907 for the predictive model. The current predictive model developed from preoperative predictors which had a good discriminative ability to differentiate between length of hospitalization less than 14 days and prolonged LOS in osteoporotic femoral neck patients. This model can be applied as ease-of use calculator application to help patients, their families and clinicians make appropriate decisions in terms of treatment planning, postoperative care program, and cost-effectiveness before patients receiving the definitive treatments.